Quand les questions en disent plus que les réponses : classification automatique des intentions dans les questions

Abstract

This work aims to automatically detect the intents of speakers in questions asked during meals. The corpus is composed of transcripts of spontaneous oral conversations. We suggest a typology of these intents based on existing work and the corpus analysis. We implement a supervised automatic classification model based on annotated data and selected linguistic features, whose results and performance are evaluated and interpreted.

Keywords

• intention detection
• dialog act
• spontaneous speech transcriptions
• questions classification
• opinion mining
• machine learning